基于优化Kriging代理模型的场景分析法求解机组组合问题  

Scenario analysis based on the optimization Kriging model for solving unit commitment problems

作　　者：崔承刚 郝慧玲 杨宁 奚培锋 CUI Chenggang;HAO Huiling;YANG Ning;XI Peifeng(School of Automation Engineering,Shanghai Electric Power University,Shanghai 200090,China;Shanghai Key Laboratory of Smart Grid Demand Response,Shanghai 200063,China)

机构地区：[1]上海电力大学自动化工程学院,上海200090 [2]上海市智能电网需求响应重点实验室,上海200063

出　　处：《电力系统保护与控制》2020年第22期49-56,共8页Power System Protection and Control

基　　金：国家自然科学基金青年科学基金项目资助(51607111);上海市科委科研计划项目资助(19DZ1205700)。

摘　　要：由于风电具有很强的波动性和不确定性,为机组组合(Unit Commitment,UC)问题带来许多问题和挑战。因此,提出了一种基于优化Kriging代理模型的场景分析法来处理风电的不确定性。首先通过"预测箱"方法生成大量场景,然后由序列优化的Kriging代理模型估计各场景所对应的经济成本。同时,根据风电不确定性及运行成本对系统的影响,采用重要性采样法削减场景。通过考虑功率平衡和风电爬坡约束的随机机组组合(Stochastic UnitCommitment,SUC)模型验证了该方法的有效性。算例分析结果表明,序列优化Kriging代理模型可以使用较少的场景预测场景运行成本。与Kantorovich距离法相比,该方法的削减结果选择了较为重要的场景,其求解结果具有更好的经济性和可靠性。Because wind power has high volatility and uncertainty,it may bring many problems and challenges to the Unit Commitment(UC)problems.Therefore,a scenario analysis method based on the sequence Kriging model is proposed to solve wind power uncertainty.It generates a large number of scenarios by a“forecast bin”method.Then the operational cost of the corresponding scenarios is estimated by a sequence optimization Kriging model.At the same time,an important sampling method is adopted to reduce scenarios given the influence of wind uncertainty combined with the operational cost.The effectiveness of this method is verified by a stochastic unit commitment model considering power balance and wind ramping constraints.It is shown that the sequence optimization Kriging model can use fewer points to estimate the operational cost of the scenario set.Compared to the Kantorovich distance method,the result of the proposed method is more representative,and the resulting solution has better economy and reliability.

关 键 词：场景分析法 序列优化Kriging代理模型 重要性采样法 机组组合 两阶段随机规划 

分 类 号：TM614[电气工程—电力系统及自动化]